Classification Of The Intertidal Saltmarsh Using Sentinel-1 And Landsat-8 Data

The tidal flat wetland possesses the ecosystem service functions of preventing the shore from erosion,supporting for biological habitat,purifying water and carbon sinks etc.Under the influence of the human-related causes such as reclamation,grazing,pollution and species intrusion as well as the global warming and the sea-level rising,the vegetation types in the system are in dynamic changing and the ecosystem is under great pressure.Monitoring the tidal flat wetland by remote sensing method is the foundation of the tidal flat wetland study and has important significance for the protection of the ecosystem and the conservation of its service functions.Subject to the high-probability recurrence of the bad weather and the incapability of penetrating the plant crown,optical remote sensing method cannot be widely adopted in the tidal flat area.The radar RS system,whose wavelength is commonly at centimeter level,can provide the ability to penetrate the cloud,fog and rain,which results in its immunity to bad weathers.Meanwhile,the longer wavelength contributes to the penetration of the plant crown,which is a favor to the identification of the vegetation types.These merits of the radar system attract more and more scholars' attentions.This article firstly discussed the back-scattering mechanism between radar signal and the plants.Using 16 Sentinel-1A VV+VH dual-pol SAR images in 2016,the temporal variation characteristics of the main land features in the southern part of the Chongming tidal flat is analyzed.Based on the analysis result,the knowledge-based decision tree was established.The Landsat 8 and Sentinel-IA images were combined to do the land cover classification of the study area.The results obtained are below:(1).The back-scattering coefficient of the Phragmites australis is greater than that of Imperate cylindrica during the leaf-off period,but the contrary is the case during the leaf-on period.It results from the different scattering mechanisms.(2)In the study area,the annual mean back-scattering intensity of the main land covers follows the relationship:Phragmites australis>Imperate cylindrica>Scirpus mariqueter/Scirpus triqueter>bare land>waterbody.The temporal standard deviation of the intensity of the Imperate cylindrica is smaller than Phragmites australis,but greater than other land features.(3)The optical image is better at distinguishing the vegetation and none-vegetation areas,while the SAR data is superior to classifying the vegetation types.A single-date SAR image will not be sufficient to classify the vegetation types accurately.During the classification of vegetation,it is necessary to consider the influence of the tide level on the scattering mechanism between the radar signal and the plants.Through band mathing of SAR images under different tidal levels will improve the classification accuracy.(4)The correlation coefficient between the back-scattering intensity of Sentinel-1A image and the NDVI is larger than the Radarsat-1 and ERS SAR image.This implies the advantage of the Sentinel-1 A data for the study about vegetation biomass.